1. Field of the Invention
The present invention relates to an image processing device, an image processing method, and a non-transitory computer readable medium recording image processing program, and more particularly, to a technique that executes an image processing by an image processing pipeline.
2. Description of Related Art
An image processing device such as a digital camera or scanner executes image processing such as a correction or noise reduction on an image acquired by shooting or scanning by the image processing pipeline.
FIG. 10 shows an example of image processing using an image processing pipeline. FIG. 10 is not a “prior art” figure because it is drafted by the inventor in order to explain a problem discovered by the inventor. The image processing pipeline shown in FIG. 10 includes a linearity correction processing circuit 91, a shading correction processing circuit 92 and a noise reduction processing circuit 93. The image processing pipeline receives a pixel value of a pixel included in an image on which image processing is to be executed. The processing circuits 91, 92 and 93 execute an image processing according to a received pixel value, and output a pixel value calculated in the image processing.
The linearity correction processing circuit 91 executes the image correction processing to recover the linearity of pixel values included in an image. The image processing unit 911 of the linearity correction processing circuit 91 multiplies a received input pixel value by a coefficient according to the input pixel value. Note that the image processing unit 911 identifies the coefficient according to a linearity correction processing specification parameter indicating that degree of the linearity correction. Further, the linearity correction processing specification parameter is a predetermined arbitrarily value. The image processing unit 911 outputs a pixel value obtained by multiplying an input pixel value by a coefficient to the shading correction processing circuit 92.
The shading correction processing circuit 92 executes the image correction processing to even out the density of an image. The image processing unit 921 of the shading correction processing circuit 92 multiplies a received input pixel value by a coefficient determined for each of the x- and y-coordinates positions of pixels included an image. Note that the image processing unit 921 identifies the coefficient according to the shading correction processing specification parameter indicating the degree of the shading correction. Further, the shading correction processing specification parameter value is a predetermined arbitrarily value. The image processing unit 921 outputs a pixel value obtained by multiplying an input pixel value by a coefficient to the noise reduction processing circuit 93.
The noise reduction processing circuit 93 executes the image processing to reduce the noise of an image. The image processing unit 931 of the noise reduction processing circuit 93 calculates a pixel value that is obtained by reducing the noise from a received input pixel by the same strength for each input pixel value. Note that the image processing unit 931 identifies the strength, by which the noise is reduced, according to a noise reduction processing specification parameter indicating the degree of the noise reduction. Further, the noise reduction processing specification parameter value is a predetermined arbitrarily value. That is, the noise reduction processing is executed on all of the pixels included an image with the same strength.
The image processing pipeline shown in FIG. 10 executes the image processing for each of the pixels included an image with different strength through the linearity correction processing and the shading correction processing. Therefore, the image generated through the linearity correction processing and the shading correction processing includes a pixel for which the noise is strengthened and a pixel for which the noise is not strengthened the noise.
However, the present inventor has found a problem that the image quality of an image generated by the noise reduction processing is degraded when the noise reduction processing is executed for all of the pixels included an image with same strength in the abovementioned image processing pipeline.
Thus, more specifically, if an image is processed with the noise reduction processing with the strength tailored to a pixel having strengthened noise, a pixel having no strengthened noise is processed with the noise reduction processing with more strength than necessary. This result in a problem that the image quality of the pixel having no strengthened noise is degraded. Then, if an image is processed with the noise reduction processing with the strength tailored to the pixel having strengthened noise, a pixel having no strengthened noise is not processed with the noise reduction processing with the necessary strength. This result in a problem that the noise remains in the pixel having strengthened noise.
As described above, the image processing pipeline does not execute image processing at the subsequent stage according to the specifics of the image processing at the preceding stage. Therefore, there is a problem that the image quality of an image generated by the image processing using the image processing pipeline is degraded.
Japanese Unexamined Patent Application Publication No. 2005-122257 discloses a technique that can simplify timing control and hardware configuration, shorten the development period, and reduce costs of an image processing device. This technique can identify an image processing corresponding to the attribute of pixel data based on tag information to identify the attribute of the pixel data, and executes the image processing on image data without instructions from the sequencer that controls the entire device. However, Japanese Unexamined Patent Application Publication No. 2005-122257 does not disclose any technique to improve the image quality of an image generated by image processing using an image processing pipeline.
Japanese Unexamined Patent Application Publication No. 2006-196992 discloses a technique that can generate history information necessary for decode processing to restore a processed image obtained by image processing to an unprocessed original input image, and then restores the processed image to the unprocessed original input image based on history information. However, Japanese Unexamined Patent Application Publication No. 2006-196992 does not disclose any technique for an image processing pipeline.